Professional Documents
Culture Documents
Copyright ? ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959, United States.
1
when quality is expressed as either the lot average for a given example, one may wish to make an estimate of the sulfur
property, or as the lot fraction not conforming to prescribed content of coal with maximum allowable error of 1 %, or e
standards. The level of a characteristic may often be taken as an = 0.01.
indication of the quality of a material. If so, an estimate of the
average value of that characteristic or of the fraction of the 7. Equations for Calculating Sample Size
observed values that do not conform to a speci?cation for that 7.1 Based on a normal distribution for the characteristic, the
characteristic becomes a measure of quality with respect to that equation for the size, n, of the sample is as follows:
characteristic. This practice is intended for use in determining 2
n 5 ~3s o/E ! (1)
the sample size required to estimate, with prescribed precision,
such a measure of the quality of a lot or process either as an where:
average value or as a fraction not conforming to a speci?ed s o = the advance estimate of the standard deviation of the
value. lot or process,
E = the maximum allowable error between the estimate to
5. Empirical Knowledge Needed be made from the sample and the result of measuring
5.1 Some empirical knowledge of the problem is desirable (by the same methods) all the units in the lot or
in advance. process, and
3 = a factor corresponding to a low probability that the
5.1.1 We may have some idea about the standard deviation
difference between the sample estimate and the result
of the characteristic.
of measuring (by the same methods) all the units in
5.1.2 If we have not had enough experience to give a precise
the lot or process is greater than E. The choice of the
estimate for the standard deviation, we may be able to state our
factor 3 is recommended for general use. With the
belief about the range or spread of the characteristic from its
factor 3, and with a lot or process standard deviation
lowest to its highest value and possibly about the shape of the
equal to the advance estimate, it is practically certain
distribution of the characteristic; for instance, we might be able
that the sampling error will not exceed E. Where a
to say whether most of the values lie at one end of the range,
lesser degree of certainty is desired a smaller factor
or are mostly in the middle, or run rather uniformly from one
may be used (Note 1).
end to the other (Section 9).
NOTE 1—For example, the factor 2 in place of 3 gives a probability of
5.2 If the aim is to estimate the fraction nonconforming,
about 45 parts in 1000 that the sampling error will exceed E. Although
then each unit can be assigned a value of 0 or 1 (conforming or
distributions met in practice may not be normal (Note 2), the following
nonconforming), and the standard deviation as well as the text table (based on the normal distribution) indicates approximate
shape of the distribution depends only on p8, the fraction probabilities:
nonconforming to the lot or process. Some rough idea con- Factor Approximate Probability of Exceeding E
cerning the size of p8 is therefore needed, which may be 3 0.003 or 3 in 1000 (practical certainty)
2.56 0.010 or 10 in 1000
derived from preliminary sampling or from previous experi-
2 0.045 or 45 in 1000
ence. 1.96 0.050 or 50 in 1000 (1 in 20)
5.3 Sketchy knowledge is sufficient to start with, although 1.64 0.100 or 100 in 1000 (1 in 10)
more knowledge permits a smaller sample size. Seldom will NOTE 2—If a lot of material has a highly asymmetric distribution in the
there be difficulty in acquiring enough information to compute characteristic measured, the factor 3 will give a different probability,
the required size of sample. A sample that is larger than the possibly much greater than 3 parts in 1000 if the sample size is small.
There are two things to do when asymmetry is suspected.
equations indicate is used in actual practice when the empirical
knowledge is sketchy to start with and when the desired 7.1.1 Probe the material with a view to discovering, for
precision is critical. example, extra-high values, or possibly spotty runs of abnor-
5.4 In any case, even with sketchy knowledge, the precision mal character, in order to approximate roughly the amount of
of the estimate made from a random sample may itself be the asymmetry for use with statistical theory and adjustment of
estimated from the sample. This estimation of the precision the sample size if necessary.
from one sample makes it possible to ?x more economically 7.1.2 Search the lot for abnormal material and segregate it
the sample size for the next sample of a similar material. In for separate treatment.
other words, information concerning the process, and the 7.2 There are some materials for which s varies approxi-
material produced thereby, accumulates and should be used. mately with μ, in which case V ( = s / μ)remains approximately
constant from large to small values of μ.
6. Precision Desired 7.2.1 For the situation of 7.2, the equation for the sample
6.1 The approximate precision desired for the estimate must size, n, is as follows:
2
be prescribed. That is, it must be decided what maximum n 5 ~3 Vo /e! (2)
deviation, E , can be tolerated between the estimate to be made
from the sample and the result that would be obtained by where:
measuring every unit in the lot or process. V o = (coefficient of variation) = s o / μ
o the advance estimate
of the coefficient of variation, expressed as a fraction
6.2 In some cases, the maximum allowable sampling error is
(or as a percentage),
expressed as a proportion, e, or a percentage, 100 e. For
e = E/ μ, the allowable sampling error expressed as a where the value of the correction factor, c 4 , depends on the
3
fraction (or as a percentage) of μ,
and size of the individual data sets (nj ) (Table 1 ).
μ = the expected value of the characteristic being mea- 8.2.2 An even simpler, and slightly less efficient estimate
sured. for s o may be computed by using the average range ( R ˉ) taken
If the relative error, e, is to be the same for all values of μ, from the several previous data sets that have the same group
then everything on the right-hand side of Eq 2 is a constant; size.
hence n is also a constant, which means that the same sample Rˉ
so 5 d (8)
size n would be required for all values of μ. 2
7.3 If the problem is to estimate the lot fraction noncon- The factor, d2 , from Table 1 is needed to convert the average
2
forming, then s o is replaced by p o so that Eq 1 becomes: range into an unbiased estimate of s o .
2
n 5 ~3/E! po ~1 2 po ! (3) 8.2.3 Example 1— Use of sˉ .
8.2.3.1 Problem — To compute the sample size needed to
where: estimate the average transverse strength of a lot of bricks when
po = the advance estimate of the lot or process fraction the desired value of E is 50 psi, and practical certainty is
nonconforming p81 and E # po desired.
7.4 When the average for the production process is not 8.2.3.2 Solution — From the data of three previous lots, the
needed, but rather the average of a particular lot is needed, then values of the estimated standard deviation were found to be
the required sample size is less than Eq 1, Eq 2, and Eq 3 215, 192, and 202 psi based on samples of 100 bricks. The
indicate. The sample size for estimating the average of the average of these three standard deviations is 203 psi. The c 4
?nite lot will be: value is essentially unity when Eq 1 gives the following
equation:
n L 5 n/[1 1 ~n/N !# (4)
n [ ~3 3 203! /50 !#2 5 2
5 149 bricks (9)
where:
n = the value computed from Eq 1, Eq 2, or Eq 3, and
for the required size of sample to give a maximum sampling
N = the lot size.
error of 50 psi, and practical certainty is desired.
This reduction in sample size is usually of little importance 8.3 For Eq 2— If s varies approximately proportionately
unless n is 10 % or more of N.
with μfor the characteristic of the material to be measured,
compute both the average, X ˉ, and the standard deviation, s, for
8. Reduction of Empirical Knowledge to a Numerical several samples that have the same size. An average of the
Value of s o (Data for Previous Samples Available) several values of v = s/ Xˉmay be used for V o .
8.1 This section illustrates the use of the equations in 8.3.1 For cases where the sample sizes are not the same, a
Section 7 when there are data for previous samples. weighted average should be used as an approximate estimate
8.2 For Eq 1— An estimate of s o can be obtained from for V o
k k
previous sets of data. The standard deviation, s, from any given 1/2
Vo 5 [ ~n j 2 1! vj / (10)
j (5 j(
~nj 2 1!#
sample is computed as: 1 5 1
where: 3
ASTM Manual on Presentation of Data and Control Chart Analysis, ASTM STP
sj = the standard deviation for sample j, 15D, 1976, Part 3, Table 27.
nj = the sample size for sample j.
8.2.1 If each of the previous data sets contains the same TABLE 1 Values of the Correction Factor C4 and d2 for Selected
A
number of measurements, nj , then a simpler, but slightly less Sample Sizes n j
efficient estimate for s o may be made by using an average ( sˉ) Sample Size 3 , (n j ) C4 d2
of the s values obtained from the several previous samples. The 2 .798 1.13
calculated sˉvalue will in general be a slightly biased estimate 4 .921 2.06
5 .940 2.33
of s o . An unbiased estimate of s o is computed as follows: 8 .965 2.85
sˉ 10 .973 3.08
so 5 c (7) A
4 As nj becomes large, C 4 approaches 1.000.
8.3.2.2 Solution — There are no data from previous samples
of this same material, but data for six samples of similar p = 21/390 = 0.054
2
materials show a wide range of resistance. However, the values n = (3/0.04) (0.054) (0.946)
of estimated standard deviation are approximately proportional = [(9 3 0.0511)/0.0016] 287.4 = 288
to the observed averages, as shown in the following text table: If the desired value of E were 0.01 the required sample size
Estimate of Coefficient would be 4600. With a lot size of 2000, equation (4) gives
Sample Avg Observed
Lot No. so = of Varia-
Sizes Cycles ˉ
range, R ˉ/3.08 A
nL = 1934 items. Although this value of nL represents about
R tion, %
1 10 90 40 13.0 14 70 % of the lot, the example illustrates the sample size required
2 10 190 100 32.5 17 to achieve the desired value of E with practical certainty.
3 10 350 140 45.5 13
4 10 450 220 71.4 16 9. Reduction of Empirical Knowledge to a Numerical
5 10 1000 360 116.9 12
Value for s o (No Data from Previous Samples of the
6 10 3550 2090 678.6 19
Avg 15.2 Same or Like Material Available)
A 9.1 This section illustrates the use of the equations in
Values of standard deviation, s, may be used instead of the estimates made
from the range, if they are preferred or already available. The use of s would be Section 7 when there are no actual observed values for the
more efficient. computation of s o .
The use of the average of the observed values of the 9.2 For Eq 1— From past experience, try to discover what
coefficient of variation for V o in Eq 2 gives the following: the smallest (a) and largest (b) values of the characteristic are
2 2 likely to be. If this is not known, obtain this information from
n 5 [ ~3 3 15.2! /10] 5 5 21.2 → 22 specimens (11)
some other source. Try to picture how the other observed
values may be distributed. A few simple observations and
for the required size of sample to give a maximum sampling questions concerning the past behavior of the process, the usual
error of 10 % of the expected value, and practical certainty is procedure of blending, mixing, stacking, storing, etc., and
desired. knowledge concerning the aging of material and the usual
8.3.2.3 If a maximum allowable error of 5 % were needed, practice of withdrawing the material (last in, ?rst out; or last in,
the required sample size would be 85 specimens. The data last out) will usually elicit sufficient information to distinguish
supplied by the prescribed sample will be useful for the study between one form of distribution and another (Fig. 1). In case
in hand and also for the next investigation of similar material. of doubt, or in case the desired precision E is a critical matter,
8.4 For Eq 3— Compute the estimated fraction nonconform- the rectangular distribution may be used. The price of the extra
ing, p, for each sample. Then for the weighted average use the protection afforded by the rectangular distribution is a larger
following equation: sample, owing to the larger standard deviation thereof.
total number nonconforming in all samples 9.2.1 The standard deviation estimated from one of the
p 5 (12) formulas of Fig. 1 as based on the largest and smallest values,
total number of units in all samples
may be used as an advance estimate of s o in Eq 1. This method
8.4.1 Example 3— Use of p: of advance estimation is acceptable and is often preferable to
8.4.1.1 Problem — To compute the size of sample needed to doubtful observed values of s, sˉ , or rˉ
.
estimate the fraction nonconforming in a lot of alloy steel track 9.2.2 Example 4— Use of s o from Fig. 1.
bolts and nuts when the desired value of E is 0.04, and practical 9.2.2.1 Problem (same as Example 1)— To compute the
certainty is desired. sample size needed to estimate the average transverse strength
8.4.1.2 Solution — The data in the following table from four of a lot of bricks when the desired value of E is 50 psi.
previous lots were used for an advance estimate of p: 9.2.2.2 Solution — From past experience the spread of values
Sample No. Non- Fraction of transverse strength for a lot of bricks has been about 1200
Low No. Size conforming onconforming psi. The values were heaped up in the middle of this band, but
1 75 3 0.040
not necessarily normally distributed.
2 100 10 0.100
3 90 4 0.044 9.2.2.3 The isosceles triangle distribution in Fig. 1 appears
4 125 4 0.032 to be most appropriate, the advance estimate s o is 1200/
Total 390 21
4.9 = 245 psi. Then
N OTE 1—What is shown here for the normal distribution is somewhat arbitrary, because the normal distribution has no ?nite endpoints.
FIG. 1 Some Types of Distributions and Their Standard Deviations
n 5 [ ~3 3 245! /50 !# 2 5 2
5 216.1 5 217 bricks (13) probability (Section 7) that the sampling error may exceed the
maximum allowable error E and to reduce the size of the
9.2.2.4 The difference in sample size between 217 and 149 sample to meet the allowable cost.
bricks (found in Example 1) is the cost of sketchy knowledge. 10.2 Eq 1 gives n in terms of a prescribed precision, but we
9.3 For Eq 2— In general, the knowledge that the use of Vo may solve it for E in terms of a given n and thus discover the
instead of s o is preferable would be obtained from the analysis precision possible for a given allowable cost that is, E = 3s o /
of actual data in which case the methods of Section 8 apply. = n . The same may be done for Eq 2 and Eq 3.
9.4 For Eq 3— From past experience, estimate approxi- 10.3 It is necessary to specify either the desired allowable
mately the band within which the fraction nonconforming is error, E, or the allowable cost; otherwise there is no proper size
2
likely to lie. Turn to Fig. 2 and read off the value ofs o = p8 of sample.
(1 - p8) for the middle of the possible range of p8and use it in
Eq 8. In case the desired precision is a critical matter, use the 11. Selection of the Sample
2
largest value of s o within the possible range of p8. 11.1 In order to make any estimate for a lot or for a process,
10. Consideration of Cost on the basis of a sample, it is necessary to select the units in the
sample at random. An acceptable procedure to ensure a random
10.1 After the required size of sample to meet a prescribed selection is the use random numbers. Lack of predictability,
precision is computed from Eq 1, Eq 2, or Eq 3, the next step such as a mechanical arm sweeping over a conveyor belt, does
is to compute the cost of testing this size of sample. If the cost not yield a random sample.
is too great, it may be possible to relax the required precision
11.2 In the use of random numbers, the material must ?rst
(or the equivalent, which is to accept an increase in the
be broken up in some manner into sampling units . Moreover,
each sampling unit must be identi?able by a serial number,
actual, or by some rule. For packaged articles, a rule is easy;
the package contains a certain number of articles in de?nite
layers, arranged in a particular way, and it is easy to devise
some system for numbering the articles. In the case of bulk
material like ore, or coal, or a barrel of bolts or nuts, the
problem of de?ning usable sampling units must take place at an
earlier stage of manufacture or in the process of moving the
materials.
11.3 It is not the purpose of this practice cover the handling
of materials, nor to ?nd ways by which one can with surety
discover the way to a satisfactory type of sampling unit.
Instead, it is assumed that a suitable sampling unit has been
de?ned and then the aim is to answer the question of how many
2
FIG. 2 Values of s , or ( s ) , Corresponding to Values of r * to draw.
ASTM International takes no position respecting the validity of any patent rights asserted in connection with any item mentioned
in this standard. Users of this standard are expressly advised that determination of the validity of any such patent rights, and the risk
of infringement of such rights, are entirely their own responsibility.
This standard is subject to revision at any time by the responsible technical committee and must be reviewed every ?ve years and
if not revised, either reapproved or withdrawn. Your comments are invited either for revision of this standard or for additional standards
and should be addressed to ASTM International Headquarters. Your comments will receive careful consideration at a meeting of the
responsible technical committee, which you may attend. If you feel that your comments have not received a fair hearing you should
make your views known to the ASTM Committee on Standards, at the address shown below.
This standard is copyrighted by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959,
United States. Individual reprints (single or multiple copies) of this standard may be obtained by contacting ASTM at the above
address or at 610-832-9585 (phone), 610-832-9555 (fax), or service@astm.org (e-mail); or through the ASTM website
(www.astm.org).